Hi All, My experience with collaborative filtering has been that it is usually based on a huge database, that you want to narrow down, and so is very specific to the architecture your database is built on / the format of your tables...
The algorithm is pretty simple though, and I'm sure you can find several examples online or in college AI textbooks. On a very basic level: 1) Start with an example of what the current user likes 2) Find other users that like the same things 3) Find things those users like but the current user hasn't seen. It's best to make some sort of correlation vector between how many things a user shares with another user, and then multiply that across the set of things the user might like, so that you could arrange them in order. If any one has other ideas, be welcome to share them. Good luck with your project. Ciao, Kevin Community Manager, Allpoetry.com "I wrote, I saw, I dove into the sea" -----Original Message----- From: Ken Williams [mailto:[EMAIL PROTECTED]] Sent: Tuesday, January 08, 2002 2:23 PM To: Leon Brocard Cc: [EMAIL PROTECTED] Subject: Re: Collaborative filtering On Tuesday, January 8, 2002, at 07:02 AM, Leon Brocard wrote: > Hullo. > > I wonder if anyone has any experience of doing collaborative filtering > in Perl (or anything else, for that matter). I haven't seen any public code for doing this, though it would certainly be a welcome project. You'd probably have to start with some of the academic research (which I'm not up-to-date on). > As I understand it, > AI::Categorize can categorize items which are similar to items that > are already categorized. That's right, it probably wouldn't be much help for collaborative filtering. -Ken
